JPH054692B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Field of Industrial Application) The present invention relates to a selection circuit that selects previously requested data at the same time as reading data stored in a storage device.

(Prior Art) As large amounts of data come to be organized as databases, processing for manipulating the databases at high speed becomes necessary. Databases have a huge amount of storage capacity, and in consideration of storage security, they are generally stored in secondary storage devices such as disks.

By the way, conventionally, in a computer system, when searching a large amount of data stored in a secondary storage device, after reading the data from the primary storage device to the primary storage device inside the computer, the central processing unit (CPU) inside the computer ) to perform the search process. However, with this method, the CPU cannot perform other processing during the search period, and all the data must be imported into the primary storage device, which increases the search time. It was hot.

Therefore, in order to solve this problem, data processing functions are added to the disk device (secondary storage device) side, and the necessary data selection process is performed during the period of reading data from the disk device. Attempts have also been made. This process of immediately processing data upon reading from a disk device and selecting data that satisfies conditions is known as on-the-fly processing.

However, in this on-the-fly processing, if the comparison speed of the comparison device that compares the key fields used for data comparison is slow, the data following the key field cannot enter the processing device after being read from the disk. , when attempting to read data that satisfies the conditions, a rotation wait occurs. In order to avoid this, the comparator is required to have a processing speed higher than the read speed of the disk. In order to improve this to some extent, it is conceivable to have a buffer in the processing device to temporarily store read data. However, even in this case, the data held in the buffer must be processed by the processing device, so if the average processing speed of the processing device is slower than the read speed of the disk, the disk will still have to handle a large amount of data. Waiting for rotation will occur.

(Problems to be Solved by the Invention) As described above, in the conventional data selection method using on-the-fly processing, high-speed processing is required for data comparison, and if this is not possible, it is necessary to wait for the rotation of the disk. There was a problem in that this caused a decrease in the read speed.

The present invention has been developed in view of the above circumstances, and its purpose is to provide a selection device that can increase the read speed without requiring a particularly high-speed comparison processing circuit.

[Structure of the Invention] (Means for Solving the Problems) The present invention provides a buffer for reading data from a storage device and retaining the data in a predetermined storage location;
The data read into the buffer provided in parallel with this buffer is input and the key field of the data is extracted, and the information of the key field is associated with the information of the storage location of the data in the buffer. an extraction means for creating and outputting an index, and a comparison processing means for inputting the index output from the extraction means and selecting the index if the key field information in the index is requested in advance. and reading means for reading out the data stored in the buffer from the buffer based on information on the storage location of the data in the selected index.

(Operation) In the present invention, data is written from the storage device to the buffer without omission. And at the same time as writing this,
The key field of the read data is extracted by the extraction means. The extraction means creates an index by associating the key field information of the extracted data with the location where the data is stored in the buffer, and outputs the index to the comparison processing means. The comparison processing means compares the key fields to determine whether the data is the previously requested data, and reads the data stored in the buffer according to the result. If the time from when data is stored in the buffer to when it is read out is shorter than the writing cycle of data to the buffer, overflow of the buffer can be avoided, so the comparison processing time of the comparison section in the comparison processing means is should be at least less than the data read cycle.

(Example) An embodiment of the present invention will be described below.

FIG. 1 is a diagram showing the configuration of a selection device according to an embodiment of the present invention.

That is, the selection device 1 according to this embodiment is provided between a system bus 2 connected to a CPU (not shown) and a secondary storage device connected thereto. The secondary storage device here consists of a disk device 3 and a disk control means 4 that controls it.

The selection device 1 includes a buffer 11, an extraction section 12, a processing section 13, and a reading section 14. That is, the disk device 3 via the disk controller 4
The data read from the buffer 11 is inputted to the buffer 11. The buffer 11 stores a certain amount of the read data. By configuring this buffer 11 with, for example, RAM (Random Access Memory), it is possible to achieve a sufficiently high read speed compared to the read speed from the disk device 3. The data input to this buffer 11 is
2 is constantly monitored. Extraction part 12
has a function of, upon detecting a key field to be compared, inputting the data of that key field, adding the leading address of the data at that time in the buffer 11 to the key field, and outputting it to the processing section 13 as an index. Processing part 1
3 compares the input key field with a predetermined standard, and if the key field satisfies the standard, outputs the start address added by the extracting unit 13 to the reading unit 14. Reading section 14
is based on the input first address information.
The data in the buffer 11 is read onto the system bus 1.

In the above configuration, it is assumed that the database stored in the disk device 3 is configured as shown in FIG. Here, one record means one unit of data read from the disk device 3 to the system bus 2, and is composed of four fields. The first field is a key field and consists of one byte. Similarly, the second field is 4 bytes,
The third field consists of 12 bytes, the fourth field consists of 12 bytes, and one record consists of 29 bytes.
Consists of bytes. Therefore, buffer 11 is 0
When records are stored starting from the address, the first address of each record will be address 0, 29, etc. starting from the first record.
The addresses will be 58 and 87. Therefore, four pieces of index data constituted by adding a start address to each key field data are shown in FIG.

Therefore, assuming that "b" is given to the processing unit 13 in advance as the key field data of the record to be selected via the system bus 2, the selection device 1 performs the following operation.

First, when the disk device 3 is activated by the disk control device 4 and the data shown in FIG. 2 is read out sequentially, these data are sequentially read into the buffer 11. When the extraction unit 12 detects the first field that is the key field of the first record, it adds the start address “0” of the data to the data “a” and outputs this as index data to the processing unit 13. do. Since the input key field data "a" is different from the requested data, the processing section does not output the start address to the reading section 14.

Next, when the extraction unit 12 detects the key field data “b” of the second record, the index data with the start address “29” added thereto is output to the processing unit 13. In the processing unit 13, since the input key field data matches the requested one, the starting address “29” is sent to the reading unit 1.
Output to 4. The reading unit 14 reads the buffer 11.
The second record stored with address 29 as the first address is read from the buffer 11 to the system bus 2. Various methods can be considered for determining the end of a record. If the record handled as in this example is a fixed-length record, the number of bytes read is fixed, so the end of the record can be easily determined. If you are dealing with variable length records, you can determine the end of the record by looking at the start address of the next record after the selected key field, or you can write the record to buffer 11. Sometimes, a tag indicating the end of the record may be added, and the end of the record may be determined when the reading unit 14 detects this tag.

Thereafter, since the key field data of the third and fourth records do not match the requested data, reading from the buffer to the system bus 2 is not performed.

By such an operation of the selection device 1, only the requested records are selected and provided to the CPU (not shown).

By the way, considering the comparison processing speed in the processing unit 13 of the selection device 1, the number of bytes in one record is 29 and the number of bytes in the key field of one record is 1, so the starting address is input to the processing unit 13 with 3 bytes. The index data to be read is 4 bytes. Therefore, disk device 3
Processing unit 1 while reading 29 bytes of data from
Since it is sufficient to process 4 bytes of data of 3, the processing speed of the processing section 13 may be 4/29=0.138 times the read speed from the disk device 3. Therefore, the processing section 13 can sufficiently perform its functions even with a circuit whose speed is much lower than the read speed of the disk.

Note that the read operation and the write operation from the disk can be performed at the same time, and the buffer 1
By appropriately setting the speed of 1 with respect to the speed of the disk device 3, it is possible to avoid inhibiting reading from the disk device 3.

It is not practical to store all the databases read from the disk device 3 in the buffer 11. This is because the database is generally expected to be too large to fit into the buffer 11. When performing this operation on a large database,
You can use batshua cyclically. In other words, once the comparison is completed, the data read into the buffer can be stored in a primary storage device (not shown) or discarded, so it can be released during the entire process. Free buffer 1 even when the database is large
This can be handled by repeatedly using one area. The leading address of the index at this time must be assigned correctly and cyclically in correspondence with the address of the buffer 11 that is used repeatedly.

In the above embodiment, the direct address of the buffer 11 is used as the index, but the information attached to the key field data may be a record identifier instead of the direct address.

FIG. 4 shows the configuration of the selection device 21 in this case, in which an address table 22 is newly added to the selection device 1 of FIG.

In this embodiment, when a key field is detected by the extraction unit 12, the identifier of the record is added to the key field data, and this is output to the processing unit 13 as an index. The relationship between the key field data and the identifier is, for example, as shown in FIG. The processing unit 13 sends the above-mentioned identifier to the reading unit 14 when the key field data matches the requested data. The reading unit 14 is
Address table 2 based on input identifier
Find the top address to be read from 2. That is, as shown in FIG. 6, the address table 22 stores correspondence information between identifiers and absolute addresses, so the reading unit 14 subtracts the start address of the record to be read from this table and fills the buffer. The target record is read from 11.

According to this embodiment, for example, when the capacity of buffer 11 is 4MB, in order to address 4MB,
Although 22 bits are required, if the number of records stored at one time in the buffer 11 is 65,536 or less, 16 bits is sufficient for the identifier.
Therefore, when using an identifier as in this example, rather than using an absolute address as information attached to key field data, the number of bits of data handled by the processing section 13 can be reduced, and the load on the processing section 13 can be reduced. can be further reduced.

In this embodiment, when reading a selected record, an operation is added in which the address table 22 indicating the correspondence between an identifier and an address must be indirectly retrieved. By performing pipeline processing on the read processing of the buffer 11, a substantial decrease in processing speed can be prevented. FIG. 7 shows a time chart of such processing. That is, in cycle 1, the processing result (identifier) of process 1 from the processing unit 13 is read. In cycle 2, the processing result of the next process 2 is read and the process 1 read in cycle 1 is read out.
Address table 2 of buffer 11 from the identifier of
An operation of subtracting 2 is performed. In cycle 3, the operation of reading the record of the address obtained from the identifier of process 1 from the buffer 11, the operation of subtracting the address from the identifier of the result of process 2, and the reading of the process result of process 3 are performed simultaneously. Then, from the next cycle 4 onwards, the records that are the selection results are successively read out. According to such processing, the read speed can be increased.

Note that the present invention is not limited to the embodiments described above. For example, in the above embodiments, the case where the secondary storage device from which the database is read is a disk device is described, but it is also applicable to the case where the storage device from which the data is read is a primary storage device. That is, the present invention is also applicable to a case where a selection process is performed on a primary storage device using the data transfer time, and the result is sent back to the primary storage device.

[Effects of the Invention] As described above, conventionally, the comparison processing in the data comparison processing section is performed immediately upon detection of a comparison location, and when the comparison processing is completed, the comparison processing section is in an idle state (transferring the subsequent data). However, according to the present invention, the data read from the storage device is stored in a buffer, and the relevant points for comparison are extracted from the data to determine the number of data to be compared. Since the comparison process is performed after reducing the amount of data, by allocating the entire reading time of the entire data to the time of the comparison process, the ratio of the number of comparison data to the number of data read into one buffer can be reduced. The speed of comparison processing can be made slower than the speed of reading data from the storage device. For this reason,
It is possible to improve the reading speed while easing the processing speed requirements of the comparison processing means.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a system using a selection device according to an embodiment of the present invention, FIG. 2 is a diagram showing the data configuration in the system, and FIG.
4 is a diagram showing the relationship between key field data extracted by the selection device and its starting address; FIG. 4 is a block diagram showing the configuration of a system using a selection device according to another embodiment of the present invention; FIG. Figure 5 is a diagram showing the relationship between key field data extracted by the selection device and its identifier, Figure 6 is a diagram showing the contents of the address table in the selection device, and Figure 7 is a processing timing diagram of the selection device. It is. 1, 21...selection device, 2...system bus,
3... Disk device, 4... Disk control device,
11...Buffer, 12...Extraction section, 13...Processing section, 14...Reading section, 22...Address table.

Claims (1)

[Scope of Claims] 1. A selection device that selects requested data from data read from a storage device, comprising: a buffer that reads data from the storage device and holds the data in a predetermined storage location; The data read into the buffer provided in parallel with this buffer is inputted, the key field of the data is extracted, and the information of the key field is associated with the information of the storage location of the data in the buffer. an extraction means for creating and outputting an index; and a comparison processing means for inputting the index output from the extraction means and selecting the index if the key field information in the index is requested in advance. and reading means for reading out the data stored in the buffer from the buffer based on information on the storage location of the data in the selected index. 2. Claim 1, wherein the information on the storage location of the data is an absolute address of the location where the data in the buffer is stored.
Selection device as described in section. 3. The information on the storage location of the data is an identifier of the data, and the reading unit knows the absolute address of the buffer corresponding to the identifier by referring to an address table. The selection device according to item 1.